Performance Characteristics of Pall Ultramet-L™ Gaskleen® Filters

Dennis Capitanio, Ph.D.

The semiconductor industry relies on state-of-the-art filters to insure that the gases that reach the wafer surface during fabrication of integrated circuits are free of particle contamination. These filters, in addition to offering very high retention levels of particles in a gas stream, must have other characteristics that insure the cleanliness of the filtered gas. The filters must not shed particles during gas pulsation and the filters must show transparency to upsets in the system such as the inadvertent introduction of moisture into the system. The filter’s transparency to the moisture in the system allows the end user to return to operating conditions quickly without waiting long periods of time for filters to desorb the moisture.

The introduction of all-metal point-of-use gas filters into the semiconductor industry has allowed the end user to operate at higher temperatures for some applications. These higher temperatures must not affect the filter’s ability to retain particles and must not contribute to particle shedding. In addition, the filter’s transparency to upsets in the gas system must also be unaffected. Pall Corporation offers the Ultramet-L(tm) Gaskleen® family of filter assemblies for the filtration of process gases in the semiconductor industry. The Ultramet-L Gaskleen 4400 series filter assembly is an all 316L stainless steel filter capable of removing particles 3 nanometers (0.003 µm) and larger. The filter assembly incorporates a sintered fiber filter pack that provides rapid gas displacement and excellent desorption characteristics. The filter assembly is manufactured and packaged under a cleanroom environment and is preconditioned for cleanliness prior to packaging. The filter assembly design coupled with preconditioning and packaging insures quick start up times after installation. Pall Corporation supplied five Ultramet-L filter assemblies (P/N GLFF4400VMM4) to American Air Liquide, Inc., Chicago Research Center for evaluation under a variety of service conditions and system upsets.

Evaluation Protocol

The scope of the testing fell under two categories: moisture testing and particle testing. The moisture testing involved three subsets which consisted of:

1. purge time of filter after removal from packaging to achieve background moisture levels
2. transparency to moisture after moisture pulse
3. transparency to moisture after moisture pulse following bakeout at 460 °C.

The particle testing involved five subsets which consisted of:

1. break-in test at 100 slpm
2. pulse test at 10 to 100 slpm
3. baking test at 460 °C
4. final purge test
5. particle retention test.

Since both the moisture and particle testing included portions before and after the bakeout of the filters, the sequence of tests varies from the list above. The material testing flow chart is shown in Figure I.

FIGURE I

Top

Test Methods and Results

Test Methods and Results

The five Pall Ultramet-L filter assemblies (P/N GLFF4400VMM4) supplied by Pall Corporation were randomly numbered 1-5 for the purpose of tracking each filter through the testing procedure. The filter assemblies can be used for extended periods of time at temperatures up to 450 °C with flows up to 50 slpm (1.76 scfm). The filter assemblies are 3.31 inches in length and supplied with 1/4 inch VCR compatible gasket seals.

The moisture and particle tests performed in this study were completed on the apparatus shown in Figure II. The moisture testing (position #1 in Figure II) used dry filtered house nitrogen at a flow of 1.5 lpm. The moisture analyzer was a Meeco Aquamatic Plus and the moisture output for the system was 2.05 ppm at the test flow. The particle shedding testing (see position #2 in Figure II) was done during the initial purge test, pulse flow test, baking tests and final purge test and used filtered nitrogen gas with flows of 10 slpm or 100 slpm as called for in the particular tests. The efficiency test (see position #3 in Figure II) used particle laden compressed air as the challenge material. The particle analyzer used for the shedding/efficiency testing was a TSI Model 3761 CNC counter with 0.01 µm (10 nanometer) sensitivity.

FIGURE II

The schematic drawing of the testing apparatus used in the particle/moisture testing

Top

Out-of Bag Moisture:

The initial purging or drydown of the filter assemblies was performed by placing each filter assembly, still in the final polyethylene bag, in a glove bag that was constantly purged by dry nitrogen. The glove bag surrounded the attachment area for the test stand and the packaged filter assembly. After the filter assembly had spent a minimum of 16 hours in the glove bag, the filter assembly was removed from the packaging and attached to the testing apparatus with a flow of 1.5 slpm. The moisture concentration passing out of the assembly was measured at 1 minute intervals until reaching 30 ppb. The 30 ppb was the background level obtained with an electropolished spool piece in place of the filter assembly. The temperature of the testing lab was regulated at 22± 1° C. The results of the drydown testing for the filter assemblies are shown in Figure III. Filter assembly #2 is not shown because of temperature drifts during the testing.

FIGURE III
The moisture drydown curves for the Pall Ultramet-L filter assemblies
after withdrawal from manufacturer packaging (Flow = 1.5 slpm)

The average drydown time for the filter assemblies was 49 minutes at 1.5 slpm to 30 ppb background. The standard deviation for the four filter assemblies was 45 minutes. American Air Liquide indicated that these results showed good and reproducible preconditioning and packaging.

Top

Moisture Transparency:

Each filter assembly was subjected to 2 ppm of moisture at 1.5 slpm for a 20 minute time period and then switched to dry nitrogen. The moisture downstream of each filter assembly was monitored during and after the moisture injection. The results of the transparency moisture testing on the filter assemblies prior to the high temperature bakeout are shown in Figure IV. Included in the Figure is a curve representing the response of a blank. The blank was an electropolished stainless steel tube that was inserted into the apparatus in place of the filter assemblies. The filters exhibited a high degree of transparency, with rapid recovery after the moisture upset.

FIGURE IV
The moisture transparency curves of the Pall Ultramet-L filter assemblies prior to bakeout
(20 minutes @ 2ppm, Flow = 1.5 slpm)

The moisture transparency curves for the filter assemblies after bakeout at 460 °C for 75 hours are shown in Figure V. The average moisture peak maximum response of the filter assemblies was 803 ppb (standard deviation = 30 ppb) before the bakeout and 666 ppb (standard deviation = 47 ppb) after the bakeout. The induction time is defined as the elapsed time between the introduction of the moisture and the first indication of the moisture exiting the filter assemblies. The average induction time before the bakeout was 10.2 minutes (standard deviation = 1 minute) and after the bakeout was 10.6 minutes (standard deviation = 1.1 minutes). The bakeout therefore had no negative effects on the transparency, demonstrating that the moisture was not retained on the stainless steel surfaces.

FIGURE V
The moisture transparency curves of the Pall Ultramet-L filter assemblies after bakeout
(20 minutes @ 2ppm, Flow = 1.5 slpm)

Top

Particle Shedding:

The particle shedding test was performed as a purge test on the filter assemblies before and after the bakeout of the assemblies. The “break in” test was performed by initially flowing 10 slpm of clean nitrogen through the filter assemblies and counting particles downstream. All assemblies showed zero particles downstream after this low flow test either before or after bakeout. The test was continued at 100 slpm (twice the recommended flow for the assemblies) and continued until the particle counter registered no counts for a six hour period. The assemblies showed particles downstream during the first hour of the test which are attributed to contamination from the moisture pulsing test. The moisture pulsing preceded this test on both the before and after bakeout portions. The particle concentration downstream of the assemblies after the first hour of purging ranged from 0 to 0.03 particles per standard cubic foot and is typical of background counts.

The particle shedding of the filter assemblies was also tested during flow pulsing. In this testing, the flow through the assemblies was varied from 10 slpm (200 seconds) to 100 slpm (160 seconds) and the particles were monitored downstream during the flow pulsing. The flow pulsing sequence and the subsequent particles counted are shown in Figure VI. The particle shedding contribution of all five assemblies during this flow pulsing test was zero, even with pulses above the maximum rated flow of the filter assemblies.

The particle shedding of the filter assemblies was monitored during temperature cycling between 50 °C and 460 °C. The flow through the filters was 10 slpm for the temperature cycling period. The filter assemblies showed no particle shedding during the temperature cycling test. The temperature cycling sequence and particle counting results are presented in Figure VII.

FIGURE VI
The particle shedding results of the Ultramet-L filter assemblies during pulsed flow testing

FIGURE VII
The temperature profile and particle shedding results during the temperature
cycling tests Particle Retention Efficiency:

Top

Particle Retention Efficiency:

The filter assemblies were tested for particle retention efficiency using compressed air which contained approximately 108 particles per cubic foot. The filter assemblies were challenged with this particulate level at flows of 50 slpm. The particles downstream of the filter assemblies were counted every minute and were equivalent to background level. The five filter assemblies showed complete retention of all challenge material. The retention efficiencies are more than sufficient in dealing with semiconductor gases with particulate loads well below the levels used in this testing.

Top

Conclusion

The Pall Ultramet-L 4400 all 316L stainless steel gas filter assembly has been tested by the Chicago Research Center of American Air Liquide under various process and upset conditions. The five samples supplied to American Air Liquide were tested for moisture contamination, moisture transparency, particle shedding and filter efficiency.

The out-of-package filter assemblies reached baseline moisture and minimal particle counts is less than one hour which ensures quick start up after installation. Moisture transparency of the filter assemblies indicated that down time will be minimal due to moisture upsets and the filter assemblies would have good transparency even after high temperature bakeout.

Flow pulsing and bakeout studies showed no particulate contribution by the assemblies during these processes. The filter assemblies showed excellent particle retention even at the upper flow limits recommended for the filter assemblies. The particle levels in the retention study were well in excess of the levels that are seen in semiconductor gases.

This study has shown good reproducibility based on random sampling of filter assemblies supplied to American Air Liquide. The study also showed that the performance of the Pall Ultramet-L 4400 filter assemblies will meet or exceed the requirements for the semiconductor industry.

Top